Studies with Comparison Groups that do Recreational Reading
We finally arrive at the gold standard, studies in which comparison groups, it is claimed, also did recreational reading. This is the sort of study that is needed to settle the issue of what aspect of AR is effective, if indeed AR is effective. Few studies have claimed to do this. Two of these studies appear in one publication, published in two versions. Vollands, Topping and Evans (1996) is an ERIC report, while Vollands, Topping and Evans (1999) is a slightly abbreviated version appearing in The Reading and Writing Quarterly. The report included two independent studies, each lasting six months. In both cases, it is claimed that AR is compared to a group that did recreational reading. It is crucial to examine them in some detail.
Vollands, Topping and Evans: Project A
This study involved very few children, especially in the comparison group (n = 12; there were 27 in the experimental group). All were 11 years old. Comparisons had 30 minutes per day of "reading time" (Vollands et. al., 1999, p. 203), but had to give "written feedback on book completion to the teacher on what they read." The fuller description in Vollands et. al. (1996, p. 53) strongly suggests that this was not merely a record of what was read but was a full "book report"; students were free to read what others had written about a book before selecting it. This means that the "written feedback" provided by students included content and most likely opinion.
The AR group in Project A had 15 minutes of reading time per day for the first five weeks of the project, which was then increased to 30 minutes. Students were also read to for 30 minutes per day, and were allowed to take AR tests on books read to them. Since it is established that readalouds have a positive effect on literacy development (Blok, 1999; Bus, van IJzendoorn, and Pellegrini, 1995), it can be argued that this read aloud time counts as exposure to reading. A factor that certainly depressed performance of the AR group was the fact that there was a severe shortage of books that the children could read and be tested on.
This study thus suffers from several problems. The comparison group sample size was small and they had to write book reports on what they read, a serious departure from pure recreational reading. Also, the AR group combined reading and read alouds: if read aloud time counts, the comparisons had less total exposure to comprehensible text. (According to Jeff McQuillan's calculations, controls had 3,600 minutes of reading. The AR group had 3,225 of reading time plus 2,850 minutes of listening to the teacher read aloud to them.)
The results are not clear. The AR group made better gains on one measure of reading comprehension (the Edinburgh, see table 9), but both groups declined on another test of reading comprehension, given only to a random subsample of the AR group. The AR group appeared to decline less, however.
Vollands et. al (1996, pp. 148-9) contains a very brief report of follow-up testing done three months after the project ended. As noted in table 9, the AR group gained 12.6 months over their previous score, and the comparisons gained 15.2 months, a substantial recovery over their decline during the treatment period. These are spectacular gains for both groups, more than four times expected growth. Vollands et. al. provide no explanation for this.